You’re right, z-masking and frustum culling can make rasterization much more efficient than having to iterate over all triangles in the scene, and those techniq

You’re right, z-masking and frustum culling can make rasterization much more efficient than having to iterate over all triangles in the scene, and those techniques are used commonly and very effectively in games. The example in that video above has everything in view though, z-masking won’t fix it or get it anywhere close to as fast as ray tracing in this specific case. The main point is that basic ray tracing has an algorithmic complexity that is logarithmic or cube root of the number of primitives (depending on your BVH), while basic rasterization has linear complexity, and that generally remains true even with many effective culling techniques. There are hybrid techniques that try to combine the best of both, so nothing here is absolute, but at some point if you develop a competitive log complexity per pixel rasterizer with a BVH, eventually it basically becomes ray tracing, and you might as well go with the method that trivially gives you shadows and reflections and bounce lighting too.